Spectral multipliers for multidimensional Bessel operators (Q650095): Difference between revisions

The paper treats the problem of spectral multipliers associated with the multidimensional Bessel operator \[ \Delta_\lambda=-\sum_{j=1}^n \Big(\frac{\partial^2}{\partial x_j^2}+\frac{2\lambda_j}{x_j}\frac{\partial}{\partial x_j}\Big), \] which is symmetric and positive in \(L^2(\mu_\lambda):=L^2((0,\infty)^n,\,\mu_\lambda)\), where \(d\mu_\lambda(x)=\prod_{j=1}^nx_j^{2\lambda_j}\,dx_j\) and \(\lambda=(\lambda_1,\ldots,\lambda_n)\in (-1\slash2,\infty)^n\) is a given type parameter. We shall use the same symbol \(\Delta_\lambda\) to denote the selfadjoint extension of the Bessel operator, given in terms of the multidimensional Hankel transform \(H_\lambda\). According to a general multiplier theorem of \textit{S. Meda } [Proc. Am. Math. Soc. 110, No.3, 639--647 (1990; Zbl 0760.42007)], the spectral multiplier \(m(\Delta_\lambda)\) is bounded on \(L^p(\mu_\lambda)\), \(1<p<\infty\), provided suitable estimates of the operator norms of the imaginary powers \(\Delta_\lambda^{i\beta}\), \(\beta\in \mathbb R\), as operators acting on \(L^p(\mu_\lambda)\), are proved with appropriate control depending on \(|\beta|\to\infty\). The authors prove the required estimates (these include also weak type (1,1) estimates) in a more general setting of multipliers of Laplace transform type. Such a multiplier is a bounded function \(m\) on \(\mathbb R_+^n\) that corresponds to a (bounded) function \(\phi\) on \(\mathbb R_+\) by the `Laplace transform type' integral formula \[ m(y)=|y|^2\int_0^\infty \exp(-t|y|^2)\phi(t)\,dt,\qquad y\in(0,\infty)^n, \] and the corresponding multiplier operator is given by \(T_\lambda^m f=H_\lambda(mH_\lambda f)\), \(f\in L^2(\mu_\lambda)\); if \(\phi=\phi_\beta(t)=\Gamma(1-i\beta)^{-1}t^{-i\beta}\), \(t\in(0,\infty)\), then \(T_\lambda^{m_\beta}=\Delta_\lambda^{i\beta} \). The estimates on \(T_\lambda^m\) for a general Laplace transform type multiplier \(m\) are stated in Corollary 1.1 and then they are specified to \(\Delta_\lambda^{i\beta}\) in Corollary 1.2; then the main result of the paper, Theorem 1.1, follows. It is worth noticing that the result of Corollary 1.1 can be seen as an extension to higher dimensions of Theorem 1.2 in \textit{J. J. Betancor, T. Martinez, L. Rodriguez-Mesa} [Can. Math. Bull. 51, No.4, 487--496 (2008; Zbl 1169.44001)], where Calderón-Zygmund operator theory was applied. In the present paper a completely different approach is used. The authors obtain a pointwise representation of \(T_\lambda^mf\) in terms of a limit of integral operators applied to \(f\) with kernels obtained by a truncation from a kernel formally associated to \(T_\lambda^m\). This is the content of Theorem 1.2 first done for \(f\in C_c^\infty((0,\infty)^n)\) and then extended to any \(f\in L^p(\mu_\lambda)\), \(1\leq p<\infty\). The extension requires an additional result stated as Theorem 1.3, where \(L^p\) and weak type (1,1) estimates for a maximal operator (based on the beforementioned truncated kernels) are proved.